We conducted a double-blind, placebo-controlled trail in which 40 patients who had undergone long-term therapy with benzodiazepines were switched to placebo or to diazepam in a dose approximately equivalent to their usual dose of the benzodiazepine; the dose of diazepam was then tapered during an eight-week period. Patients were assessed clinically and psychologically and had weekly sessions of behavioral therapy. The subjects who received placebo had more symptoms, assessed their symptoms as more severe, and stopped taking the study drug at a higher rate than those receiving the tapering doses of diazepam. The subjects in the placebo group also had symptoms shortly after being switched to placebo, whereas those in the diazepam group had symptoms much later. Some withdrawal symptoms were distinct from those of anxiety (e.g., tinnitus, involuntary movement, and perceptual changes). Withdrawal symptoms occurred earlier in patients who had received short-acting benzodiazepines than in those who had received long-acting benzodiazepines. Symptoms gradually disappeared over a four-week period in both the placebo and the diazepam groups. Serial determination of plasma benzodiazepine concentrations was a useful way to assess compliance, treatment outcome, and relapse during withdrawal. We conclude that a clinically important, mild, but distinct withdrawal syndrome occurs after discontinuation of long-term therapeutic use of benzodiazepines.